Flexilevel Adaptive Testing Paradigm: Validation in Technical Training.
Abstract
This study was designed to empirically assess a computerized adaptive testing model in an ongoing technical training system. The model was a modification of Lord's flexilevel paradigm and consisted of: the sequencing of test items in a difficulty hierarchy, adaptive entry of students into the test at a difficulty level appropriate to their predicted score, and systematic movement of students up and down the hierarchy based upon their performance on preceding items. The subjects were 444 airmen enrolled in the Inventory Management/Materiel Facilities Course at Lowry Air Force Base, Colorado. They participated in the study as part of the normal achievement testing requirement for Block II of the course. Predictor variables for individualized entry were three reading tests administered to students prior to course enrollment. A within-subject design was employed in comparing the adaptive strategy to conventional testing on dependent variables or performance, test reliability, and test time. This involved first administering items according to the flexilevel algorithm and then, after the student exited from the test at either the top or bottom end of the hierarchy, presenting all remaining items. Consequently, both adaptive and conventional test scores were obtained for each subject. The results revealed a high positive part-whole correlation between flexilevel and total test scores. Internal consistency indices for the two forms were essentially equivalent. With regard to length and time, however, the flexilevel test required nine less items than the entire test, yielding a length reduction of 39.5 percent with a concomitant time savings of 18.4 percent.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1977
- Accession Number
- ADA042977
Entities
People
- Dickie A. Harris
- Duncan N. Hansen
- Steven Ross
Organizations
- University of Memphis